Using machine learning to estimate atmospheric Ambrosia pollen concentrations in Tulsa, OK.

Area of Science:

• Environmental science
• Aerobiology
• Data science

Background:

• Airborne pollen, particularly from Ambrosia species, is a significant allergen impacting public health.
• Accurate estimation of daily pollen abundance is crucial for allergy management and epidemiological studies.
• Traditional methods for pollen monitoring can be labor-intensive and spatially limited.

Purpose of the Study:

• To develop and evaluate machine learning models for estimating daily airborne Ambrosia pollen abundance in Tulsa, OK.
• To compare the performance of different machine learning algorithms, including LASSO, neural networks, and random forests.
• To identify key meteorological and land surface variables influencing Ambrosia pollen levels.

Main Methods:

• Utilized 27 years of historical pollen observations.
• Integrated 85 meteorological and land surface variables as input features.
• Employed machine learning algorithms: Least Absolute Shrinkage and Selection Operator (LASSO), neural networks, and random forests.
• Evaluated model performance based on prediction accuracy.

Main Results:

• Random forests demonstrated the highest performance in estimating daily Ambrosia pollen abundance.
• The study identified significant relationships between environmental variables and pollen counts.
• Machine learning models successfully integrated diverse datasets for pollen estimation.

Conclusions:

• Machine learning, particularly random forests, offers a powerful approach for accurate airborne pollen abundance estimation.
• The developed models can enhance pollen forecasting and allergy management strategies.
• Further research can explore incorporating additional data sources and refining model parameters for broader applicability.